Docling Document Reader
Spring AI provides a DocumentReader API to build an ingestion pipeline for loading, parsing, chunking, and storing documents to be used as context for scenarios such as Retrieval Augmented Generation (RAG) flows.
Arconia offers an implementation of the DocumentReader API (DoclingDocumentReader) that uses Docling as the engine for processing documents and making them ready to be stored in a vector database. It’s built on top of Arconia Docling, which provides seamless integration with Docling, a powerful AI-powered document conversion service that transforms documents into structured formats.
Quick Start
Let’s see how you can get started with the Docling Document Reader in your Spring AI application.
Dependencies
First, include the Docling Document Reader dependency in your project.
-
Gradle
-
Maven
dependencies {
implementation 'io.arconia:arconia-ai-docling-document-reader'
}<dependency>
<groupId>io.arconia</groupId>
<artifactId>arconia-ai-docling-document-reader</artifactId>
</dependency>Arconia publishes a BOM (Bill of Materials) that you can use to manage the version of the Arconia libraries. While not required, it is recommended to use the BOM to ensure that all dependencies are compatible.
-
Gradle
-
Maven
dependencyManagement {
imports {
mavenBom "io.arconia:arconia-bom:0.20.0"
}
}<dependencyManagement>
<dependencies>
<dependency>
<groupId>io.arconia</groupId>
<artifactId>arconia-bom</artifactId>
<version>0.20.0</version>
<type>pom</type>
<scope>import</scope>
</dependency>
</dependencies>
</dependencyManagement>Dev Services
Arconia Dev Services provide zero-code integrations for services your application depends on, both at development and test time, relying on the power of Testcontainers and Spring Boot.
When working with Docling Document Reader, you can use the Docling Dev Service to automatically start a Docling Serve instance during development and testing, giving you the possibility to parse documents without manually setting up a Docling Serve instance.
To enable the Docling Dev Service, add the following dependency to your project:
-
Gradle
-
Maven
dependencies {
testAndDevelopmentOnly "io.arconia:arconia-dev-services-docling"
}<dependency>
<groupId>io.arconia</groupId>
<artifactId>arconia-dev-services-docling</artifactId>
<scope>runtime</scope>
<optional>true</optional>
</dependency>
<build>
<plugins>
<plugin>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-maven-plugin</artifactId>
<configuration>
<includeOptional>false</includeOptional>
</configuration>
</plugin>
</plugins>
</build>Ingesting Documents with Docling
Given a story.pdf file, you can build a data ingestion pipeline as follows:
@Component
public class IngestionPipeline {
private final DoclingServeApi doclingServeApi;
private final VectorStore vectorStore;
public IngestionPipeline(DoclingServeApi doclingServeApi, VectorStore vectorStore) {
this.doclingServeApi = doclingServeApi;
this.vectorStore = vectorStore;
}
@PostConstruct
void run() {
Resource file = new ClassPathResource("story.pdf");
List<Document> documents = DoclingDocumentReader.builder()
.doclingServeApi(doclingServeApi)
.files(file)
.build()
.get();
vectorStore.add(documents);
}
}The DoclingDocumentReader will use Docling to process the file, transform it into smaller chunks, and build a resulting collection of Spring AI’s Documents. Finally, you can use Spring AI’s VectorStore API to convert the documents into embeddings and store them into a vector database.
| For more information on how to build a data ingestion pipeline in Spring AI, check out the dedicated documentation. |